RESULTS

Richness and Endemism

There is a distinct trend in richness from high at low latitudes to low at higher latitudes (Fig. 6). When richness is regressed against latitude, R2=0.58 (Fig. 7). Seven regions at middle latitudes are somewhat distant re lative to the trend line. When those seven (WP, TORR, PILB, BULL, BARK, LEB, and MDB) are removed, the R2 value for richness versus latitude increases to 0.83 (Fig. 7).

The regions with highest endemism included all western, central and southern regions (Fig. 8). An "endemic line" between areas of "high" and "low" endemism effectively splits Australia into two halves. NEQ was the only region with relatively high end emism east of the line. The number of regions in which each species occurs (Fig. 9) shows seventy-eight species (47% of the fauna) in only one region, sixty-four in two to seven regions, sixteen in eight through fifteen regions, and nine in between sixte en and twenty-two regions.

Similarity Coefficients

All four similarity coefficients produced similar UPGMA trees (Fig. 10). NTSYS warned of tied tries during calculations; however, the three trees identified as having ties were identical for each similarity coefficient. Cophenetic co rrelations were all between 0.92 and 0.95. Jaccard's and Dice's coefficients produced identically branched trees, while Ochiai's Coefficient varied only in the position of WP. Kulczynski's #2 Coefficient was different in the positions of PILB, WP, and M DB, and the cluster LEB, BULL, and BARK differed in internal arrangement.

Ball clusters in consensus trees of single- and complete-link trees of Jaccard's and Dice's coefficients were identical (Fig. 11). Ochiai's differed only in the position of W and EKIM (Fig. 11). Kulczynski's #2 differed in positions and composition o f the upper cluster, MDB and SAG clustered, and BARK did not cluster with LEB and BULL (Fig. 11).

Non-metric Multidimensiona l Scaling

Results were similar for both Jaccard's and Kulczynski's #2 coefficients (Fig. 12). Stress values were 0.12 and 0.11 respectively and cophenetic correlations were a high 0.90 and 0.96.

Parsimony Analysis

Parsimony analysis found fifteen trees with length 214 (CI=0.411, RI=0.753, RL=0.310, and HI=0.589). Strict consensus produced only two polytomies (Fig. 13). Bootstrap values were mostly low, only 6 branches having values > 50%. This is due to the small number of characters (an average of 2.8) relative to the number of OTUs. Decay index values were also low, maximum values being three.